Have ge_set_gej_var, gej_double_var and ge_set_all_gej_var initialize all fields of their outputs.

[roconnor-blockstream]

Previous behaviour would not initialize r->x and r->y values in the case where infinity is passed in.

[gmaxwell]

utACK dd6c3de

[real-or-random]

Concept ACK

Unless I'm mistaken, I think VG_UNDEFs here would be helpful. See #889 (comment).

[gmaxwell]

This function exits without leaving anything uninitialized, so nothing to vg_undef here. Aside, in the past I hadn't UNDEFed anywhere in the codebase itself, only in the tests to avoid injecting vg instrumentation in the production binary.. but we do that now for the CT tests and I think were were satisfied it was benign enough.

edit: oh I see, the set infinity could be undefing x/y. I suppose it could be. But that should be a separate PR if it gets done.

[real-or-random]

edit: oh I see, the set infinity could be undefing x/y. I suppose it could be.

Right, that's what I had in mind.

But that should be a separate PR if it gets done.

Fair point.

[real-or-random]

utACK dd6c3de

[gmaxwell]

utACK 45b6468

[real-or-random]

@roconnor-blockstream You could cherry pick the top two commits from https://github.com/real-or-random/secp256k1/commits/20210504_ge_set_gej_var

edit: I have verified that the new test fails on master.

[real-or-random]

Ah damn, this was not the latest version on my branch -.- Let me push the right one. Sorry.

[real-or-random]

Now my branch should have the right commits... I somehow had lost them. (The reflog is nice sometimes. :))

[gmaxwell]

ACK 14c9739

[sipa]

utACK 14c9739

[real-or-random]

ACK 14c9739